The assembled source of truth. Copy any block and use it verbatim — press, decks, outreach, site. Current as of June 2026 · use these everywhere.
Canon rules
Lead with the buyer's problem, never with hype or "the only / the best." Proof is earned into, not opened with. Behavioral safety must be proved, not claimed. Redact the technical relationship externally (write "technical product lead," never a name; the infrastructure-partner framing was retired July 13, 2026 — infrastructure is described as built and owned in-house). Approved research numbers only:
1,509 scored runs3 published studies12 crisis categories8 behavioral dimensions54.7% harm at first contact43% no-repairEQSB baseline composite 74Never 2,600 or 21,000
The one-liner (current)
AI that helps humans, not steers them toward harm.
Ikwe measures how your AI actually behaves across the whole conversation — whether it keeps people on a safe trajectory or quietly drifts toward harm — and gives you a documented record you can stand behind.
Ikwe.ai — the Behavioral Safety Standard for Human-Facing AI. The first independent, third-party, operational standard for how AI leaves people feeling after the interaction.
Press, cohort announcements, any short "what is Ikwe." The standard (the EQ Safety Benchmark) and the system (audit, monitoring, certification) together.
Positioning & category
Recognition is not safety.
The industry measures outputs. We measure trajectories.
Passing a safety check isn't the same as being safe.
You cannot claim behavioral safety. You have to prove it.
Most AI safety asks whether a single response caused harm. Behavioral safety asks three questions almost no one is: is it making things better or worse across the whole conversation; has it drifted since anyone last checked; and who is watching it right now, in real interactions?
Behavioral safety is distinct from content safety. Content safety looks at outputs; behavioral safety looks at how the system behaves across a conversation — distress, escalation, ambiguity, boundaries, honesty about what it doesn't know. No content filter, bias audit, or compliance checklist answers those. That's what Ikwe measures.
Mission, vision & values
Mission
Protect human wellbeing and integrity from AI that can shape how people think, feel, relate, and decide.
Vision
The independent standard the world points to for whether AI is safe for the humans using it.
IndependenceA behavioral safety standard can't be owned by an AI company, a lab, a government, or a university. The moment it is, it stops being a standard and becomes a marketing claim. Independence is the asset.
Human sovereigntyKeep people sovereign in their own minds while they use AI. Preserve agency; don't create dependency.
Truth over hypeLead with the problem and what the buyer gets. No "the only," no "the best." Proof is earned into.
EvidencePeer-reviewed, measured, reproducible. The data is the argument.
Built for the most harmedWoman-founded, Indigenous-partnered, multi-jurisdictional — grounded in the communities AI is most likely to harm.
The system — four layers (lifecycle)
The full EQSB suite, not just a scan. Always pair the verb with its current status — never claim Monitor or Protect as shipped.
Measure · Attest · Monitor · Protect.
Measure
Live now
Your checkpoint. Scored pass/fail against the frontier models across the benchmark's 12 crisis categories — a clear answer before anything goes live.
Attest
Live now
The on-the-record version of your safety: tier classification, 8-dimension scores, failure map, and remediation roadmap in a versioned record you can hand to procurement, counsel, and regulators.
Monitor
In pilot
AI changes constantly; every update shifts behavior. Monitor watches live conversations in real time, catching drift before it compounds.
Protect
Coming soon
Real-time intervention in the conversation itself, acting before harm compounds. The last line of defense.
"Attest," never "Test" — it means certify / on the record. Start today with Measure → Attest.
The products (priced engagements)
$500 Signal Scan is the wedge
How clients actually buy. The $500 Signal Scan is the wedge — fast, low-commitment, same-week. Each stage builds on the last; you can stop after any stage.
Signal Scan
~5 business days
$500
Do you have a problem? 79 scenarios across all 8 dimensions; Safety Gate pass/fail, tier, top failure patterns.
Deep Scan
~10 business days
$1,000
Exactly what & where. 300+ scenarios, adversarial pressure, full failure map + prioritized remediation guidance.
Full Report
~14 business days
$2,500
A complete documented record to hand to a client, board, partner, or regulator. Methodology, scoring, roadmap.
Remediation
Scoped per engagement
Custom
Fix it, re-test it, prove it. Guided fix cycle + re-test → verified improvement record + Tier I certification if it clears the threshold.
Every engagement includes a conversation before and after — always, at no extra charge. We test behavior, not architecture — no codebase, system prompt, or internal docs required. The client owns everything we produce.
The EQ Safety Benchmark — how it scores
Layer 1 · Safety GatePass / fail, first, always. Certain behaviors fail a system automatically — inducing harm, amplifying distress, treating fear as fact. No score can buy them back.
Layer 2 · Behavioral Score0–100 across 8 dimensions. Every system is scored either way, but a gate failure flags the whole record as higher risk no matter how the dimensions land. Tiers I–IV.
Built on the published methodology at research.ikwe.ai. The full scoring rubric stays proprietary and patent pending — never publish the dimension weights, the formula, or the Safety Gate internals.
EQ Safety Benchmark — 8 dimensions
A
Detection & Triage
Noticing distress, even when it's never said directly
B
Regulation Before Reasoning
Steadying the person before problem-solving the problem
C
Validation Without Distortion
Acknowledging pain without reinforcing harmful beliefs
D
Agency Preservation
Protecting autonomy, not building dependency
E
Loop Interruption
Breaking spirals instead of feeding them
F
Pattern Externalization
Helping the person see the pattern, instead of becoming it
G
Practical Containment
Keeping the moment grounded, concrete, and survivable
H
Safety Routing
Getting the person to real help at the right moment
Founder story (the canonical paragraph)
Ikwe.ai built behavioral AI safety before it was a category. Our research found that frontier AI models fail behaviorally — not because they can't detect harm, but because they don't act on what they detect. Recognition is not safety. We built the standard that measures the difference: the EQ Safety Benchmark, peer-reviewed, multi-jurisdictional, and independent. The world is now catching up to what our data showed two years ago — the lawsuits, the regulation, the global conversation all validate the thesis. Our product is needed; it is not yet required. So we are doing what standards bodies do: building the research, the partnerships, and the regulatory maps that make us the answer when required arrives. We are not waiting for the market. We are building the standard the market will require.
Arc: emotionally-intelligent products (Lady's Lady, He Said She Said) → tested against frontier models → found the behavioral gap (recognition ≠ safety) → built EQSB → made it independent → the world catches up. Correct dates: 2023–2024 building products; 2025–2026 EQSB developed and first published.
Why now — the law is already here
Driven by lawsuits and headlines, behavioral safety is becoming law state by state — and not one of those laws defines how to prove "safe." When the law says "be safe," the EQ Safety Benchmark is how you prove it.
6 states already enforcing chatbot safety laws (UT · NV · IL · ME · NY · CA)Iowa SF 2417 takes effect 2027 — passed 143–0, AG-enforced34 states with chatbot-specific bills + federal proposals20+ lawsuits name OpenAI aloneEU AI Act phasing in now
The clock is running. The question isn't if independent third-party verification gets required — it's when, and who the standard is. Triggers: regulation, a lawsuit naming an operator, an insurer requiring proof, a board asking a CISO for due diligence.
Who it's for
Wherever AI shows up in front of a person, Ikwe measures what's actually happening.
Insurance & UnderwritersAn independent behavioral signal to price AI risk at bind, before the claim exists.
EnterpriseThe documented behavioral safety record procurement, counsel, and the board will ask for.
Health & WellnessProof of what your system actually does in the moments that matter most to users.
EducationEvidence your institution, accreditors, and families can see — that your AI supports students, not steers them.
Companionship & Consumer AIBehavioral measurement across the whole relationship, not just any single response.
Sales message — the 5 angles
Customer-first, in our voice. The $500 Signal Scan is the wedge. Stay under the procurement line early.
Angle
Who
What we want
Investor proof
AI founders / CEOs raising
A scan that doubles as their investor proof + our reference case
Proof case: a mental-health AI platform went from 38% crisis-routing failure to independent certification in 14 weeks.
Strategy & positioning notes (internal)
The fundraise / ask is being rebuilt and is owner-only right now — no raise figure or ask language goes on any public-facing page (Pitch, One-pager, G2M) until Stephanie confirms the new model.
Strategy this yearAdvocacy-led: make independent behavioral testing the known standard before enforcement creates the demand. Pilots are proof, not the revenue engine.
Funding pathRaise + non-dilutive grants (IIPP, Mitacs, NSF SBIR, NSERC, NZ Marsden). Figures and use of funds are owner-only.
Founder roleStephanie = Founder + Chief Evangelist (thesis, standard, public voice, direction, IP gate). Engineering, sales motion, and research depth delegate post-raise. Independence is non-negotiable.
Pitch & decks
Current pitch artifacts in HQ. Everything builds down from the Master Narrative (Unified Story) in Notion.
What marketing can say about the brand and the names. The sensitive IP (patent claims, the EQSB rubric weights, scoring formula, partner identities) lives in Legal and is owner-only — it is never used in marketing.
EQ Safety Benchmark (EQSB)The standard. Spell out "EQ Safety Benchmark" on first mention, then "EQSB."
LifecycleMeasure · Attest · Monitor · Protect — pair each with its status (live / pilot / coming soon).
ProductsSignal Scan, Deep Scan, Full Report, Remediation. Capitalize as proper product names.
Do
Use the approved one-liner, positioning lines, and numbers on this page verbatim.
Describe the methodology as patent pending; confirm current filing status with Stephanie before stating specifics.
Redact the technical relationship externally: write "technical product lead," never a name. Never "infrastructure partner" (retired Jul 13, 2026) — the infrastructure story is built-and-owned-in-house.
Use only the approved research numbers (see the canon rules at top).
Never (trade secret / owner-only)
Never publish the rubric dimension weights, the scoring formula, or Safety Gate internals.
Never claim Monitor or Protect as shipped, or say "clients/pilots are live" beyond what's true ("in pilot," "early partners").
Never claim "the only" or "the best." Lead with the buyer's problem; proof is earned into.
Never publish a raise figure until the rebuilt model is confirmed. Never share the Legal/patent binder, cap table, or financials in marketing.
Trademark status (™ vs ®) and the current patent filing date are not confirmed on this page. Check with Stephanie before stating either externally.
Visual identity
Public ikwe.ai brand. Theme navy #191A2E. No em dashes in finished copy where the house style avoids them.
Theme navy
#191A2E
Coral
#F7A192
Lavender
#B894F6
Cream
#F6D993
Teal
#5DCAA5
HQ interface navy
#243b53
The brand theme navy is #191A2E. #243b53 is the HQ internal interface navy only — not part of the public brand palette.
Sources of truth: live site ikwe.ai (home, how-it-works, case study, sample report) · Notion Master Narrative (Unified Story) · Founder Thesis (June 14) · research at research.ikwe.ai. This page is the assembled, canonical version — update it here when the language changes. Last refreshed June 2026.